The NVIDIA HGX H100 8-GPU board is a high-performance computing platform designed for large-scale AI training and inference workloads. It features eight H100 Tensor Core GPUs interconnected via third-generation NVSwitch technology to provide massive throughput and low-latency communication.
- Architecture: NVIDIA Hopper
- Form Factor: SXM5 8-GPU Baseboard
- GPU Count: 8x NVIDIA H100 Tensor Core GPUs
- Interconnect: NVIDIA NVLink and NVSwitch (900 GB/s per GPU)
- Memory: Up to 640GB HBM3 total (80GB per GPU)
- Performance: Up to 32 petaFLOPS FP8 precision
- TDP: Up to 700W per GPU (system level cooling required)
- Interface: PCIe Gen5 support
- Application: Large Language Model (LLM) training, deep learning, and scientific simulation
Related Products
A relatable product